Fast and efficient dense variational stereo on GPU
|
|
- Aileen Erica Walsh
- 7 years ago
- Views:
Transcription
1 Fast and efficient dense variational stereo on GPU Julien Mairal, Renaud Keriven and Alexandre Chariot CERTIS ENPC Marne-la-Vallee cedex 2 France julien.mairal@m4x.org, keriven@certis.enpc.fr, chariot@certis.enpc.fr Abstract Thanks to their high performance and programmability, the latest graphics cards can now be used for scientific purpose. They are indeed very efficient parallel Single Instruction Multiple Data (SIMD) machines. This new trend is called General Purpose computation on Graphics Processing Unit (GPGPU [4]). Regarding the stereo problem, variational methods based on deformable models provide dense, smooth and accurate results. Nevertheless, they prove to be slower than usual disparity-based approaches. In this paper, we present a dense stereo algorithm, handling occlusions, using three cameras as inputs and entirely implemented on a Graphics Processing Unit (GPU). Experimental speedups prove that our approach is efficient and perfectly adapted to the GPU, leading to nearly video frame rate reconstruction. 1. Introduction For the last ten years graphics cards have grown significantly in terms of performance, functionalities and relative importance in a computer. Compared to Central Processing Units (CPUs), the growth rate has been far more higher and the GPUs do not obviously follow Moore s law. The newest NVIDIA/ATI cards confirm this trend in terms of pipelines numbers (24/48), memory amounts (512MB), clock frequencies (650MHz), memory bandwidth (50GB/sec), etc. The idea of using them for something else than threedimensional rendering emerged when GPU designers made them programmable, in order to provide more expressiveness to software developers who wanted to design their own effects. Nevertheless, being provided with a programmable GPU does not mean having any generic parallel machine. A graphics card is mainly designed to render a scene and display it. Therefore, very tight constraints remain. In a simplified view, data structures in video memory are either a set of vertices or a texture. A vertex is a set of coordinates in the three-dimensional space, augmented with attributes, such as texture coordinates or colors. A texture is an array of pixel color values, each usually composed of four channels when using the RGBA mode. Moreover, because of the programming paradigm, namely the Concurrent Read Exclusive Write Single Instruction Multiple Data (CREW SIMD) programming model, only specific algorithms can be adapted and efficiently implemented on a GPU. In this paper, we present an implementation of a dense stereo algorithm which runs entirely on a GPU. It is based on the variational framework of deformable models proposed by the pioneering work of [5]. In this work, the authors consider the n-view stereovision problem and recover the entire surface of an object, minimizing some energy that incorporates both photometric consistency and regularizing constraints. Here, for speed reasons, we restrict ourselves to the case of two or three cameras, and model the surface as a depth map from one reference camera. Although providing smooth and accurate results for three-dimensional shape reconstruction, such approaches are usually neglected in everyday stereo-vision applications because of their relative lack of efficiency. This is what motivated our work. Indeed, most of the efforts that have been made toward efficient stereo, are disparity-based. Instead of recovering three-dimensional information directly, the process is a two stages one: first, estimating the correspondences (disparity) between two rectified images; second reconstructing the three-dimensional object. In this context, the authors of [6, 12, 7, 15, 11, 14, 13] have developed real-time or near real-time stereo algorithms on GPUs. For example, the algorithm designed by Yang and Pollefeys in [13] can reach up 289 million disparity evaluations per second on a (now old!) ATI Radeon Exactly like the state of the art methods classified on the reference Middlebury Stereo Vision Page [10], this impressive result is not focused toward accurate threedimensional shape reconstruction. Our goal here is differ-
2 ent: it consists in obtaining as fast as possible a coherent and accurate surface, from two or three images or video streams. A first attempt to use GPUs for the same problem has been made by the authors of [16]. Yet, although they present visibly good results, they use a non-robust difference of gray levels as a photometric consistency criterion where we use instead normalized cross-correlation. Above all, the main difference between our work and theirs is that we use a mathematically founded gradient descent where they let their surface move at constant speed, just looking for a decrease of energy. Finally, note that, although correlation based, we do neither require any image rectification nor need the images be oriented the same way: the images are back-projected on the surface and correlated directly on it. To improve convergence and cope with local minima, our algorithm is multi-scale. Camera selection and occlusion are taken into account through considerations based on the normal to the surface. 2. GPU programming 2.1. Overview of GPU architecture We describe here in a simplified manner the graphics pipeline. For more details, we refer the reader to [8, 1, 4, 2]. Figure 1 presents a view of a graphics card: 1. Vertices are sent to the card with directives indicating their topology (full or wireframe polygons, triangles or quads, etc.). 2. The vertex units, which are Multiple Instructions Multiple Data (MIMD) processors on the latest cards, compute, for each vertex, a position on the screen (or on an off-screen buffer), as well as some new attributes (see [8]). This first step is now programmable. Vertex programs are also called vertex shaders. 3. Some operations are then performed, such as culling (discarding vertices which are out of a previously defined bounding box) or stencil testing (discarding vertices which are on a previously defined area). 4. The so-called rasterizer converts then the vertices and their topology into sets of pixels on the screen. It links the output registers of the vertex units to the input registers of the pixel units so that each pixel receives interpolated values from the three vertices of the triangle it belongs to. 5. The pixel units, which are Single Instruction Multiple Data (SIMD) processors compute the color of each pixel. This step is programmable, pixels programs being also called fragment shaders, and is usually the Figure 1. A simplified view of a recent graphics pipeline one considered for GPGPU, although the increasing programmability of vertex shaders now turns then into good GPGPU candidates in some cases. 6. The graphics card memory is mainly organized into textures that can be seen as two-dimensional arrays. Textures are used to communicate between the CPU and the GPU or between two different successive computations (called rendering for obvious reasons). Pixels programs have a full random read access to textures and exclusive write access to one location in one 1 texture (the previously mentioned CREW model). Vertex programs can access textures too, but this access is very slow Iterative mesh deformation To take advantage of the efficiency of GPUs, it is essential to avoid as much as possible data transfers between the GPU and the CPU: they are very slow compared to the internal GPU video memory access. In the context of an iterative process like the one we will have to deal with, there are two GPU/CPU transfer issues: 1 Actually, up to four textures using so-called Multiple Rendering Target (MRT)
3 1. An iteration should be able to process some input data into output data available for the next iteration without transmitting output back to the CPU. When using the pixels units only, data inputs and output are stored into so-called Render Textures (or more recently Frame Buffer Objects (FBO)), allowing the card to render into an off-screen buffer, that can be linked afterward to an input texture register. This is the usually adopted solution. When using also the vertex units, data inputs must be stored into Vertex Buffer Objects (VBO), that are actually a set of vertices data into the video memory. It is possible to transfer data from pixel textures to a VBO. This is the solution we adopted: our surface is handled through a so-called Vertex Array, vertices attributes being stored in a VBO. O2 y x O secondary camera main camera image 2 image 1 z d_m = OM M deformable surface S deformation direction triangle T Figure 2. Framework of the cameras. 2. One needs a way to test some stopping criterion, despite the fact that the pixels do not communicate, without transmitting data back to the CPU. This is hopefully possible thanks to a mechanism called occlusion query, which counts the number of rendered pixels. It is then easy to have every single pixel artificially not rendered is some test is true, and being instantaneously warned if this test is true for all the pixels. 3. Two cameras we use a more robust energy, based on normalized crosscorrelation, where the images are back-projected onto the surface (or the plane tangent to S around m) and correlated on some neighborhood of m: E(S) = (1 ρ(i 1 Π 1, I 2 Π 2, m))ds(m) (1) S We choose a discrete version of E(S), computing the correlations on each triangle only. Denoting by T the triangles of S, we thus take: 3.1. Model E(S) = T E T = T (1 ρ T (I 1 Π 1, I 2 Π 2 )) (2) Let us consider the case where two cameras are used. Dealing with a third camera is similar and will be detailed further. Considering two fully-calibrated cameras and choosing camera 1 as a reference, we model the object as a regular triangulated deformable surface S (see figure 2) where each vertex M lie on a fixed ray issued from the optical center O of camera 1. Although asymmetrical, this representation spares a lot of computation. Let d M be the distance between point O and point M. We will design some energy E(S) that is actually a function of the d M s and minimize it by means of a gradient descent Energy The simplest energy could be the sum of the differences between the gray levels: E(S) = (I 2 Π 2 (m) I 1 Π 1 (m)) 2 ds(m) S where the I i are the images, Π i the projections associated to the cameras and denotes the function composition. Because the sum is done all over the surface, one could think this measure is robust enough, like mentioned in [16]. We have tried it without success on real images. Following [5], with (omitting dependencies in T in the notations): ρ T = < I 1, I 2 > I 1. I 2 1 < I 1, I 2 > = (I 1 Π 1 (m) I 1 Π 1 ) I i Π i = A T 1 A T T T I i = < I i, I i > (I 2 Π 2 (m) I 2 Π 2 )dm I i Π i (m)dm A T denoting the surface of triangle T. Note that we do not use any surface metric anymore. Multiplying the correlation by the area of the triangle would lead to the smallest possible surface (i.e. converging toward point O). This behavior is common to all active contour based methods and is usually solved with some balloon force. We could have done this here but we got better results with above formulation added with some regularization term (see further) Gradient We minimize the energy by means of a gradient descent, using a multi-resolution scheme, in terms of both mesh and (3)
4 image sizes, in order to cope with local minima as much as possible. E(S) being actually a function of the distances d M s, we have to compute its gradient with respect to these distances. Let V (M) be the set of triangles to which a vertex M belongs, we write: E(S) = T V (M) E T (4) Computing these quantities from equations (2) and (3) when M is one of the vertices of T is straightforward but gives rather long expressions we will not detail here. We refer the reader to [9] for their complete writing. To discuss their GPU implementation, the important fact is that they boil down to double sums (the mean values being encapsulated inside the main sum) depending on the following quantities: I i Π i (m) (5) where point m is a point of T and M one of its vertices GPU discretization Implementing the above quantities on a GPU, the continuous sums F (m)dm involved in equation (2) T for some F will indeed be replaced by discrete sums p F (p)dm(p) where p is determined by the rasterizer and each pixel shader will compute f(p). Let T be the triangle (M 1, M 2, M 3 ), p will be some barycenter p = α 1 M 1 +α 2 M 2 +α 3 M 3 with α i 0 and α 1 +α 2 +α 3 = 1. We then need to compute the quantities of equation (5), i.e. D i,k (α 1, α 2, α 3 ) = I i Π i (p) k for k = 1, 2, 3 and every (α 1, α 2, α 3 ) chosen by the rasterizer. A direct computation involving camera projections gives: D i,k (α 1, α 2, α 3 ) = g i (α k f i (M k ), Π i (p), ( I i ) Π i (p)) where f i and g i are simple geometric function given in [9]. Note that, when a pixel shader is called for a point p, the value α k is not known. Yet, it is possible to get the required α k f i (M k ): adding a virtual attribute to the vertices and setting it to f i (M k ) for M k and to 0 for M j (j k), automatically provides the interpolated quantity α k f i (M k ) when processing p. Moreover, the rasterizer can compute Π i (p) because Π i (p) = α 1 Π i (M 1 ) + α 2 Π i (M 2 ) + α 3 Π i (M 3 ). An advantage of this is that the quantities depending on the vertices M k are computed once only by the vertex units and not for each point p. Note also that, when rendering a given triangle T = (M 1, M 2, M 3 ), the three values D i,1, D i,2 and D i,3 can be computed simultaneously for each p thanks to vectorial capacities of the GPU. Now, choosing to render the surface from camera 1 point of view, we get that, for a given p T, the value of I 1 Π 1 (p) does not depend on the positions of the vertices of T, i.e: 3.5. Summations D 1,k (α 1, α 2, α 3 ) = 0 Each F (p) being computed for some F by pixel shaders for each p T, we still need to recover p T F (p)dm(p). Such a reduction is a classical problem on SIMD machines and logarithmic complexity algorithms are usually designed. Here, our triangles have a small number of pixels thanks to our multi-scale approach adapting mesh size to image dimensions. Thus a simple pass where each pixel shader deals with one triangle and performs a loop 2 over it points is much more efficient. We also have to sum, for each vertex, quantities over its related triangles (equation (2)). Again, vertices are now assigned to pixel shaders that perform a loop over their respective triangles Regularization As we mentioned previously, we left any surface metric apart and should add manually a regularization term to the energy. Mean curvature motion could be used here (or in an equivalent manner adding the area of the surface to the energy). Actually, we got better results directly smoothing d M, adding the following term to the gradient: ( 1 ) K ( d M ) d M N(M) M N(M) where N(M) is the set of the neighbors of M and K some constant adapted to the considered level of detail Stopping criterion An obvious stopping term based on the maximum value of the gradient gives good results. In fact, a fixed number of iterations for each level of detail gives faster convergence, yet keeping similar results Complete scheme As depicted figure 3, one complete iteration is finally: 1. Assign the surface vertices to the GPU vertices and render, computing D 2,k (p), (k = 1, 2, 3) for every point p. 2 available on recent GPUs
5 2. Assign the surface triangles to pixel shaders and perform the double summation yielding E T. 3. Assign the surface vertices to pixel shaders and perform the neighborhood summation leading to the gradient E(S). Update the vertices position d M according to this gradient. 4. Update the VBO containing the vertices positions and optionnaly compute a stopping criterion Results Step 1 True surface rasterization. Pixel processing. for each pixel m of image 2, equation (5) is calculated (except for diagonals) Pixel texture Our experiments were done on a standard 3GHz PC with a (now outdated!) NVIDIA Geforce 7800 GTX 256 graphics card. The OpenGL library and the Cg language were used. I 2, I 1 and I 2 at the different resolutions were preprocessed in a first step. The textures were in a 16 bits mode (yielding 40% faster programs). We obtained a fast convergence with an average 10 iterations per level of detail. Our multi-resolution approach proved to prevent from converging toward a local minimum. As a reference, we developed a CPU version minimizing the same energy than the GPU version. This version was cautiously written and compiled with the latest compilers with all of the optimizations turned on. Table 1 shows the speedups between CPU and GPU versions. For each level of detail, the image and mesh resolutions are given. We observed a mean speedup of 10 times to 15 times. This shows that our implementation makes the most of the graphics pipeline. Some results and total running times are given on figures 7, 8, 9 and 10. For each presented result, we ran about 8 iterations for each level of detail mentioned table 1 except for the two last level of detail that are run respectively 4 and 2 times. A total running time of about 250ms is observed for global convergence. As expected, our algorithm is not as fast as those dealing with disparity maps. Again, we insist that our targeted applications are different. A drawback so far is that we do not handle occlusions. We could easily prevent the algorithm taking occluded triangles into account. We will do it show in the next section when using a third camera. 4. Three cameras - Occlusions Our model can easily be extended to three cameras. Let us denote them C, R and L, respectively a center, a right and a left cameras. C will be the reference camera (the role played previously by camera 1) and the correlation will be computed between cameras C and R or between cameras C and L. Note that the cameras do not have to be Pixel processing : one pixel per true surface vertex. The regularization term is added, the gradient descent is performed. A new position for each vertex is calculated. Step 2 Pixel processing : (one pixel per triangle) Calculation of the new derivatives of the energy for each triangle using results of step 1 in a texture. (note : A MRT is used.) Step 3 Step 4 Pixel texture Pixel texture The vertex Array is updated. Data from the pixel texture is copied to the VBO. Step 4 bis (optional) A Stop test is performed thanks to an Occlusion Query Figure 3. Example of one iteration for a mesh of 3x3 vertices and an image of 8x8 pixels. Actually, our implementation begins with a 3x3 mesh and a 32x32 image. On this figure, a red quad in dotted line represents one rendered pixel. The plain lines represent the mesh.
6 Image Mesh GPU CPU Speedup khz 555 Hz khz 116 Hz Hz 28.6 Hz Hz 7.5 Hz Hz 7.3 Hz Hz 7.2 Hz 9.0 Table 1. Iterations per seconds for each level of detail. Figure ms, two cameras. Figure 4. First and second data sets. Figure ms, two cameras. Figure 5. Third and fourth data sets (courtesy of Pr. Kyros Kutulakos (University of Toronto)). Figure ms, two cameras. Figure 6. A data set for three cameras from [3]. Figure ms, two cameras.
7 disposed in a (left, center, right) way. The only criteria to assign the cameras are that: (i) the surface is modeled from the optical center of C, and (ii) correlation between L and R is not taken into account. This choice being done, we divide at each iteration the triangles into two sets S R and S L, S R (respectively S L ) being the set of the triangles that will be correlated between camera C and camera R (respectively camera L). The associated energy is indeed similar to the one given equation (2): E(S) = T S R (1 ρ T (I C Π C, I R Π R )) + T S L (1 ρ T (I C Π C, I L Π L )) In our tests, we simply use the normal to the triangles to choose the best secondary camera, R or L. Moreover, occlusions can easily be taken into account. Rendering the surface from secondary cameras viewpoints, we determine whether a triangle is seen or not from them. When a triangle is not seen by any secondary camera, we just assign it neither to S R nor to S L. Then, keeping the energy given by equation (6), a triangle that is not seen by at least C and another camera does not contribute to the energy. Actually, we ignore also triangles that are not enough in front of the cameras in a certain sense. This yields a first way to handle discontinuities, an important point that our model does not take into account so far. We have implemented this algorithm on both CPU and GPU. The CPU version does not take occlusions into account (a CPU visibility test would be too slow) and runs only 9% slower than its two cameras version. Figure 11. Two cameras with occluded areas To implement the GPU version, we use a so-called Stencil Buffer to classify the triangles into two groups, being then able to work on one group only at a time. This functionality consist in discarding the rendering of a pixel p when the value of the stencil buffer at p is below, equal to or greater than some reference value. The solution is thus to memorize in the stencil buffer whether a pixel (or a triangle) belongs to S R, to S L or to none of them. This Figure 12. Three cameras. test is hardware implemented and is therefore very fast. An overview of the algorithm for the GPU version is given figure 13. Note that rendering pixels in two passes (first S R and then S L ) could lead to memory cache inefficiency if the pixels were randomly distributed between S R and S L. Hopefully, this is not the case here! Our first results show a 30% overhead between the two and three cameras GPU versions. We are still in the process of designing a more optimized three cameras GPU version. A first result is presented figures 11 and 12. Thanks to occlusion handling, but also thanks to more precision where the three cameras are available, the three cameras reconstruction (fig. 12) is more correct and more accurate than the two cameras one (fig. 11). 5. Videos We have tested our algorithm on video sequences. Convergence on the first frame takes about 250ms. Yet, taking advantage of temporal continuity, we just take the surface recovered at time t as an initial value for convergence at time t + 1. Some preliminary tests, essentially consisting in adjusting choices of levels of detail and number of iterations, showed promising results at a rate of 8 frames per second for the two cameras algorithm. With the recently available graphics cards a 12 frames/sec rate should already be possible. Moreover, it should be noted that our approach could easily be turned into a parallel version using two graphics cards with a minimum of communication at each iteration, yielding full video rate for a very affordable price. 6. Conclusion In this paper, we have presented a fast dense stereo algorithm based on variational principles. Handling occlusions, it takes two or three cameras as inputs and is entirely implemented on Graphics Processing Units. Experiments show that it is efficient and well adapted to the GPU: speedup of about 10 to 15 times are coherent with the graphics card we
8 Step 1 Classification of the triangles. Results are written in the stencil buffer. Step 2 used. The reconstructed surface is accurate. This fully justifies considering this approach instead of usual disparitybased algorithms for certain applications. Taking advantage of temporal continuity, we achieved reconstruction at a video rate of about 8 frames/sec. Future work includes taking discontinuities into account, dealing with more than one GPU, and investigating the reconstruction of a complete object several such systems. The Stencil Buffer for the steps 3 and 4 is written. References Step 7 Step 3 A Stencil test is used to handle one group of triangles. equation (5) is computed for each pixel. Step 4 A Stencil test is used to handle the other group. equation (5) is computed for each pixel. Step 5 The derivative of the energy is computed for one group of triangles using a Stencil test. Step 6 The derivative of the energy is computed for the other group of triangles using a Stencil test. Regularization and gradient descent. The new positions of the vertices are calculated. Step 8 The vertex Array is updated after some iterations, the level of detail is changed. Figure 13. The GPU implementation with three cameras. [1] Opengl specifications. [2] [3] quan/webpami/ pami.html. [4] [5] O. D. Faugeras and R. Keriven. Variational principles, surface evolution, pdes, level set methods, and the stereo problem. IEEE Transactions on Image Processing, 7(3): , [6] I. Geys, T. P. Koninckx, and L. V. Gool. Fast interpolated cameras by combining a gpu based plane sweep with a maxflow regularisation algorithm. In 3DPVT 04: Proceedings of the 3D Data Processing, Visualization, and Transmission, 2nd International Symposium on (3DPVT 04), pages , Washington, DC, USA, IEEE Computer Society. [7] M. Gong and Y.-H. Yang. Near real-time reliable stereo matching using programmable graphics hardware. In CVPR 05: Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 05) - Volume 1, pages , Washington, DC, USA, IEEE Computer Society. [8] E. Kilgariff and R. Fernando. The GeForce 6 Series GPU Architecture, volume GPU Gems 2, chapter Chap [9] J. Mairal and R. Keriven. A gpu implementation of variational stereo. Technical Report O5-13, CERTIS, November [10] D. Scharstein and R. Szeliski. A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. Int. J. Comput. Vision, 47(1-3):7 42, [11] J. Woetzel and R. Koch. Multi-camera real-time depth estimation with discontinuity handling on pc graphics hardware, August [12] J. Woetzel and R. Koch. Real-time multi-stereo depth estimation on gpu with approximative discontinuity handling, March [13] R. Yang and M. Pollefeys. A versatile stereo implementation on commodity graphics hardware. Real-Time Imaging, 11(1):7 18, [14] R. Yang, M. Pollefeys, H. Yang, and G. Welch. A unified approach to real-time, multi-resolution, multi-baseline 2d view synthesis and 3d depth estimation using commodity graphics hardware. [15] C. Zach, K. Karner, and H. Bischof. Hierarchical disparity estimation with programmable 3d hardware, [16] C. Zach, A. Klaus, M. Hadwiger, and K. Karner. Accurate dense stereo reconstruction using graphics hardware, 2003.
GPU(Graphics Processing Unit) with a Focus on Nvidia GeForce 6 Series. By: Binesh Tuladhar Clay Smith
GPU(Graphics Processing Unit) with a Focus on Nvidia GeForce 6 Series By: Binesh Tuladhar Clay Smith Overview History of GPU s GPU Definition Classical Graphics Pipeline Geforce 6 Series Architecture Vertex
More informationGPU Architecture. Michael Doggett ATI
GPU Architecture Michael Doggett ATI GPU Architecture RADEON X1800/X1900 Microsoft s XBOX360 Xenos GPU GPU research areas ATI - Driving the Visual Experience Everywhere Products from cell phones to super
More informationL20: GPU Architecture and Models
L20: GPU Architecture and Models scribe(s): Abdul Khalifa 20.1 Overview GPUs (Graphics Processing Units) are large parallel structure of processing cores capable of rendering graphics efficiently on displays.
More informationA Prototype For Eye-Gaze Corrected
A Prototype For Eye-Gaze Corrected Video Chat on Graphics Hardware Maarten Dumont, Steven Maesen, Sammy Rogmans and Philippe Bekaert Introduction Traditional webcam video chat: No eye contact. No extensive
More informationComputer Graphics Hardware An Overview
Computer Graphics Hardware An Overview Graphics System Monitor Input devices CPU/Memory GPU Raster Graphics System Raster: An array of picture elements Based on raster-scan TV technology The screen (and
More informationGraphics Cards and Graphics Processing Units. Ben Johnstone Russ Martin November 15, 2011
Graphics Cards and Graphics Processing Units Ben Johnstone Russ Martin November 15, 2011 Contents Graphics Processing Units (GPUs) Graphics Pipeline Architectures 8800-GTX200 Fermi Cayman Performance Analysis
More informationGPGPU Computing. Yong Cao
GPGPU Computing Yong Cao Why Graphics Card? It s powerful! A quiet trend Copyright 2009 by Yong Cao Why Graphics Card? It s powerful! Processor Processing Units FLOPs per Unit Clock Speed Processing Power
More informationThe Evolution of Computer Graphics. SVP, Content & Technology, NVIDIA
The Evolution of Computer Graphics Tony Tamasi SVP, Content & Technology, NVIDIA Graphics Make great images intricate shapes complex optical effects seamless motion Make them fast invent clever techniques
More informationIntroduction GPU Hardware GPU Computing Today GPU Computing Example Outlook Summary. GPU Computing. Numerical Simulation - from Models to Software
GPU Computing Numerical Simulation - from Models to Software Andreas Barthels JASS 2009, Course 2, St. Petersburg, Russia Prof. Dr. Sergey Y. Slavyanov St. Petersburg State University Prof. Dr. Thomas
More informationIntroduction to Computer Graphics
Introduction to Computer Graphics Torsten Möller TASC 8021 778-782-2215 torsten@sfu.ca www.cs.sfu.ca/~torsten Today What is computer graphics? Contents of this course Syllabus Overview of course topics
More informationImage Processing and Computer Graphics. Rendering Pipeline. Matthias Teschner. Computer Science Department University of Freiburg
Image Processing and Computer Graphics Rendering Pipeline Matthias Teschner Computer Science Department University of Freiburg Outline introduction rendering pipeline vertex processing primitive processing
More informationOverview Motivation and applications Challenges. Dynamic Volume Computation and Visualization on the GPU. GPU feature requests Conclusions
Module 4: Beyond Static Scalar Fields Dynamic Volume Computation and Visualization on the GPU Visualization and Computer Graphics Group University of California, Davis Overview Motivation and applications
More informationRecent Advances and Future Trends in Graphics Hardware. Michael Doggett Architect November 23, 2005
Recent Advances and Future Trends in Graphics Hardware Michael Doggett Architect November 23, 2005 Overview XBOX360 GPU : Xenos Rendering performance GPU architecture Unified shader Memory Export Texture/Vertex
More informationRadeon HD 2900 and Geometry Generation. Michael Doggett
Radeon HD 2900 and Geometry Generation Michael Doggett September 11, 2007 Overview Introduction to 3D Graphics Radeon 2900 Starting Point Requirements Top level Pipeline Blocks from top to bottom Command
More informationA Short Introduction to Computer Graphics
A Short Introduction to Computer Graphics Frédo Durand MIT Laboratory for Computer Science 1 Introduction Chapter I: Basics Although computer graphics is a vast field that encompasses almost any graphical
More informationIntroduction to GPGPU. Tiziano Diamanti t.diamanti@cineca.it
t.diamanti@cineca.it Agenda From GPUs to GPGPUs GPGPU architecture CUDA programming model Perspective projection Vectors that connect the vanishing point to every point of the 3D model will intersecate
More informationAnnouncements. Active stereo with structured light. Project structured light patterns onto the object
Announcements Active stereo with structured light Project 3 extension: Wednesday at noon Final project proposal extension: Friday at noon > consult with Steve, Rick, and/or Ian now! Project 2 artifact
More informationQCD as a Video Game?
QCD as a Video Game? Sándor D. Katz Eötvös University Budapest in collaboration with Győző Egri, Zoltán Fodor, Christian Hoelbling Dániel Nógrádi, Kálmán Szabó Outline 1. Introduction 2. GPU architecture
More information1. INTRODUCTION Graphics 2
1. INTRODUCTION Graphics 2 06-02408 Level 3 10 credits in Semester 2 Professor Aleš Leonardis Slides by Professor Ela Claridge What is computer graphics? The art of 3D graphics is the art of fooling the
More informationThe Scientific Data Mining Process
Chapter 4 The Scientific Data Mining Process When I use a word, Humpty Dumpty said, in rather a scornful tone, it means just what I choose it to mean neither more nor less. Lewis Carroll [87, p. 214] In
More informationAdvanced Rendering for Engineering & Styling
Advanced Rendering for Engineering & Styling Prof. B.Brüderlin Brüderlin,, M Heyer 3Dinteractive GmbH & TU-Ilmenau, Germany SGI VizDays 2005, Rüsselsheim Demands in Engineering & Styling Engineering: :
More informationReal-time Visual Tracker by Stream Processing
Real-time Visual Tracker by Stream Processing Simultaneous and Fast 3D Tracking of Multiple Faces in Video Sequences by Using a Particle Filter Oscar Mateo Lozano & Kuzahiro Otsuka presented by Piotr Rudol
More informationRadeon GPU Architecture and the Radeon 4800 series. Michael Doggett Graphics Architecture Group June 27, 2008
Radeon GPU Architecture and the series Michael Doggett Graphics Architecture Group June 27, 2008 Graphics Processing Units Introduction GPU research 2 GPU Evolution GPU started as a triangle rasterizer
More informationCSE 564: Visualization. GPU Programming (First Steps) GPU Generations. Klaus Mueller. Computer Science Department Stony Brook University
GPU Generations CSE 564: Visualization GPU Programming (First Steps) Klaus Mueller Computer Science Department Stony Brook University For the labs, 4th generation is desirable Graphics Hardware Pipeline
More informationultra fast SOM using CUDA
ultra fast SOM using CUDA SOM (Self-Organizing Map) is one of the most popular artificial neural network algorithms in the unsupervised learning category. Sijo Mathew Preetha Joy Sibi Rajendra Manoj A
More informationA Binary Adaptable Window SoC Architecture for a StereoVision Based Depth Field Processor
A Binary Adaptable Window SoC Architecture for a StereoVision Based Depth Field Processor Andy Motten, Luc Claesen Expertise Centre for Digital Media Hasselt University tul IBBT Wetenschapspark 2, 50 Diepenbeek,
More informationInteractive Level-Set Deformation On the GPU
Interactive Level-Set Deformation On the GPU Institute for Data Analysis and Visualization University of California, Davis Problem Statement Goal Interactive system for deformable surface manipulation
More informationMonash University Clayton s School of Information Technology CSE3313 Computer Graphics Sample Exam Questions 2007
Monash University Clayton s School of Information Technology CSE3313 Computer Graphics Questions 2007 INSTRUCTIONS: Answer all questions. Spend approximately 1 minute per mark. Question 1 30 Marks Total
More informationDynamic Resolution Rendering
Dynamic Resolution Rendering Doug Binks Introduction The resolution selection screen has been one of the defining aspects of PC gaming since the birth of games. In this whitepaper and the accompanying
More informationSilverlight for Windows Embedded Graphics and Rendering Pipeline 1
Silverlight for Windows Embedded Graphics and Rendering Pipeline 1 Silverlight for Windows Embedded Graphics and Rendering Pipeline Windows Embedded Compact 7 Technical Article Writers: David Franklin,
More informationHardware-Aware Analysis and. Presentation Date: Sep 15 th 2009 Chrissie C. Cui
Hardware-Aware Analysis and Optimization of Stable Fluids Presentation Date: Sep 15 th 2009 Chrissie C. Cui Outline Introduction Highlights Flop and Bandwidth Analysis Mehrstellen Schemes Advection Caching
More informationOpenGL Performance Tuning
OpenGL Performance Tuning Evan Hart ATI Pipeline slides courtesy John Spitzer - NVIDIA Overview What to look for in tuning How it relates to the graphics pipeline Modern areas of interest Vertex Buffer
More informationA NEW METHOD OF STORAGE AND VISUALIZATION FOR MASSIVE POINT CLOUD DATASET
22nd CIPA Symposium, October 11-15, 2009, Kyoto, Japan A NEW METHOD OF STORAGE AND VISUALIZATION FOR MASSIVE POINT CLOUD DATASET Zhiqiang Du*, Qiaoxiong Li State Key Laboratory of Information Engineering
More informationImage Synthesis. Transparency. computer graphics & visualization
Image Synthesis Transparency Inter-Object realism Covers different kinds of interactions between objects Increasing realism in the scene Relationships between objects easier to understand Shadows, Reflections,
More informationTexture Cache Approximation on GPUs
Texture Cache Approximation on GPUs Mark Sutherland Joshua San Miguel Natalie Enright Jerger {suther68,enright}@ece.utoronto.ca, joshua.sanmiguel@mail.utoronto.ca 1 Our Contribution GPU Core Cache Cache
More informationScan-Line Fill. Scan-Line Algorithm. Sort by scan line Fill each span vertex order generated by vertex list
Scan-Line Fill Can also fill by maintaining a data structure of all intersections of polygons with scan lines Sort by scan line Fill each span vertex order generated by vertex list desired order Scan-Line
More informationSegmentation of building models from dense 3D point-clouds
Segmentation of building models from dense 3D point-clouds Joachim Bauer, Konrad Karner, Konrad Schindler, Andreas Klaus, Christopher Zach VRVis Research Center for Virtual Reality and Visualization, Institute
More informationReal-Time Realistic Rendering. Michael Doggett Docent Department of Computer Science Lund university
Real-Time Realistic Rendering Michael Doggett Docent Department of Computer Science Lund university 30-5-2011 Visually realistic goal force[d] us to completely rethink the entire rendering process. Cook
More informationImplementation of Canny Edge Detector of color images on CELL/B.E. Architecture.
Implementation of Canny Edge Detector of color images on CELL/B.E. Architecture. Chirag Gupta,Sumod Mohan K cgupta@clemson.edu, sumodm@clemson.edu Abstract In this project we propose a method to improve
More informationLecture 11: Multi-Core and GPU. Multithreading. Integration of multiple processor cores on a single chip.
Lecture 11: Multi-Core and GPU Multi-core computers Multithreading GPUs General Purpose GPUs Zebo Peng, IDA, LiTH 1 Multi-Core System Integration of multiple processor cores on a single chip. To provide
More informationINTRODUCTION TO RENDERING TECHNIQUES
INTRODUCTION TO RENDERING TECHNIQUES 22 Mar. 212 Yanir Kleiman What is 3D Graphics? Why 3D? Draw one frame at a time Model only once X 24 frames per second Color / texture only once 15, frames for a feature
More informationVisualization and Feature Extraction, FLOW Spring School 2016 Prof. Dr. Tino Weinkauf. Flow Visualization. Image-Based Methods (integration-based)
Visualization and Feature Extraction, FLOW Spring School 2016 Prof. Dr. Tino Weinkauf Flow Visualization Image-Based Methods (integration-based) Spot Noise (Jarke van Wijk, Siggraph 1991) Flow Visualization:
More informationClassification of Fingerprints. Sarat C. Dass Department of Statistics & Probability
Classification of Fingerprints Sarat C. Dass Department of Statistics & Probability Fingerprint Classification Fingerprint classification is a coarse level partitioning of a fingerprint database into smaller
More informationWriting Applications for the GPU Using the RapidMind Development Platform
Writing Applications for the GPU Using the RapidMind Development Platform Contents Introduction... 1 Graphics Processing Units... 1 RapidMind Development Platform... 2 Writing RapidMind Enabled Applications...
More informationEfficient Storage, Compression and Transmission
Efficient Storage, Compression and Transmission of Complex 3D Models context & problem definition general framework & classification our new algorithm applications for digital documents Mesh Decimation
More informationA Computer Vision System on a Chip: a case study from the automotive domain
A Computer Vision System on a Chip: a case study from the automotive domain Gideon P. Stein Elchanan Rushinek Gaby Hayun Amnon Shashua Mobileye Vision Technologies Ltd. Hebrew University Jerusalem, Israel
More informationLecture Notes, CEng 477
Computer Graphics Hardware and Software Lecture Notes, CEng 477 What is Computer Graphics? Different things in different contexts: pictures, scenes that are generated by a computer. tools used to make
More informationInteractive Level-Set Segmentation on the GPU
Interactive Level-Set Segmentation on the GPU Problem Statement Goal Interactive system for deformable surface manipulation Level-sets Challenges Deformation is slow Deformation is hard to control Solution
More informationTemplate-based Eye and Mouth Detection for 3D Video Conferencing
Template-based Eye and Mouth Detection for 3D Video Conferencing Jürgen Rurainsky and Peter Eisert Fraunhofer Institute for Telecommunications - Heinrich-Hertz-Institute, Image Processing Department, Einsteinufer
More informationInteger Computation of Image Orthorectification for High Speed Throughput
Integer Computation of Image Orthorectification for High Speed Throughput Paul Sundlie Joseph French Eric Balster Abstract This paper presents an integer-based approach to the orthorectification of aerial
More informationCUDA programming on NVIDIA GPUs
p. 1/21 on NVIDIA GPUs Mike Giles mike.giles@maths.ox.ac.uk Oxford University Mathematical Institute Oxford-Man Institute for Quantitative Finance Oxford eresearch Centre p. 2/21 Overview hardware view
More informationVision based Vehicle Tracking using a high angle camera
Vision based Vehicle Tracking using a high angle camera Raúl Ignacio Ramos García Dule Shu gramos@clemson.edu dshu@clemson.edu Abstract A vehicle tracking and grouping algorithm is presented in this work
More informationIP Video Rendering Basics
CohuHD offers a broad line of High Definition network based cameras, positioning systems and VMS solutions designed for the performance requirements associated with critical infrastructure applications.
More informationKapitel 12. 3D Television Based on a Stereoscopic View Synthesis Approach
Kapitel 12 3D Television Based on a Stereoscopic View Synthesis Approach DIBR (Depth-Image-Based Rendering) approach 3D content generation DIBR from non-video-rate depth stream Autostereoscopic displays
More informationMetropoGIS: A City Modeling System DI Dr. Konrad KARNER, DI Andreas KLAUS, DI Joachim BAUER, DI Christopher ZACH
MetropoGIS: A City Modeling System DI Dr. Konrad KARNER, DI Andreas KLAUS, DI Joachim BAUER, DI Christopher ZACH VRVis Research Center for Virtual Reality and Visualization, Virtual Habitat, Inffeldgasse
More informationImproved Real-Time Stereo on Commodity Graphics Hardware
Improved Real-Time Stereo on Commodity Graphics Hardware Ruigang Yang Marc Pollefeys Sifang Li Department of Computer Science Department of Computer Science Department of Computer Science University of
More informationCUBE-MAP DATA STRUCTURE FOR INTERACTIVE GLOBAL ILLUMINATION COMPUTATION IN DYNAMIC DIFFUSE ENVIRONMENTS
ICCVG 2002 Zakopane, 25-29 Sept. 2002 Rafal Mantiuk (1,2), Sumanta Pattanaik (1), Karol Myszkowski (3) (1) University of Central Florida, USA, (2) Technical University of Szczecin, Poland, (3) Max- Planck-Institut
More informationOptimizing AAA Games for Mobile Platforms
Optimizing AAA Games for Mobile Platforms Niklas Smedberg Senior Engine Programmer, Epic Games Who Am I A.k.a. Smedis Epic Games, Unreal Engine 15 years in the industry 30 years of programming C64 demo
More informationConfiguring Memory on the HP Business Desktop dx5150
Configuring Memory on the HP Business Desktop dx5150 Abstract... 2 Glossary of Terms... 2 Introduction... 2 Main Memory Configuration... 3 Single-channel vs. Dual-channel... 3 Memory Type and Speed...
More informationOptimizing Unity Games for Mobile Platforms. Angelo Theodorou Software Engineer Unite 2013, 28 th -30 th August
Optimizing Unity Games for Mobile Platforms Angelo Theodorou Software Engineer Unite 2013, 28 th -30 th August Agenda Introduction The author and ARM Preliminary knowledge Unity Pro, OpenGL ES 3.0 Identify
More informationInteractive Information Visualization using Graphics Hardware Študentská vedecká konferencia 2006
FAKULTA MATEMATIKY, FYZIKY A INFORMATIKY UNIVERZITY KOMENSKHO V BRATISLAVE Katedra aplikovanej informatiky Interactive Information Visualization using Graphics Hardware Študentská vedecká konferencia 2006
More informationColor correction in 3D environments Nicholas Blackhawk
Color correction in 3D environments Nicholas Blackhawk Abstract In 3D display technologies, as reviewers will say, color quality is often a factor. Depending on the type of display, either professional
More informationParallel Simplification of Large Meshes on PC Clusters
Parallel Simplification of Large Meshes on PC Clusters Hua Xiong, Xiaohong Jiang, Yaping Zhang, Jiaoying Shi State Key Lab of CAD&CG, College of Computer Science Zhejiang University Hangzhou, China April
More informationBuilding an Advanced Invariant Real-Time Human Tracking System
UDC 004.41 Building an Advanced Invariant Real-Time Human Tracking System Fayez Idris 1, Mazen Abu_Zaher 2, Rashad J. Rasras 3, and Ibrahiem M. M. El Emary 4 1 School of Informatics and Computing, German-Jordanian
More informationGPGPU: General-Purpose Computation on GPUs
GPGPU: General-Purpose Computation on GPUs Randy Fernando NVIDIA Developer Technology Group (Original Slides Courtesy of Mark Harris) Why GPGPU? The GPU has evolved into an extremely flexible and powerful
More informationBUILDING TELEPRESENCE SYSTEMS: Translating Science Fiction Ideas into Reality
BUILDING TELEPRESENCE SYSTEMS: Translating Science Fiction Ideas into Reality Henry Fuchs University of North Carolina at Chapel Hill (USA) and NSF Science and Technology Center for Computer Graphics and
More informationIntroduction to Cloud Computing
Introduction to Cloud Computing Parallel Processing I 15 319, spring 2010 7 th Lecture, Feb 2 nd Majd F. Sakr Lecture Motivation Concurrency and why? Different flavors of parallel computing Get the basic
More informationCanny Edge Detection
Canny Edge Detection 09gr820 March 23, 2009 1 Introduction The purpose of edge detection in general is to significantly reduce the amount of data in an image, while preserving the structural properties
More informationPHOTOGRAMMETRIC TECHNIQUES FOR MEASUREMENTS IN WOODWORKING INDUSTRY
PHOTOGRAMMETRIC TECHNIQUES FOR MEASUREMENTS IN WOODWORKING INDUSTRY V. Knyaz a, *, Yu. Visilter, S. Zheltov a State Research Institute for Aviation System (GosNIIAS), 7, Victorenko str., Moscow, Russia
More information3D Computer Games History and Technology
3D Computer Games History and Technology VRVis Research Center http://www.vrvis.at Lecture Outline Overview of the last 10-15 15 years A look at seminal 3D computer games Most important techniques employed
More informationTouchstone -A Fresh Approach to Multimedia for the PC
Touchstone -A Fresh Approach to Multimedia for the PC Emmett Kilgariff Martin Randall Silicon Engineering, Inc Presentation Outline Touchstone Background Chipset Overview Sprite Chip Tiler Chip Compressed
More informationSpace Perception and Binocular Vision
Space Perception and Binocular Vision Space Perception Monocular Cues to Three-Dimensional Space Binocular Vision and Stereopsis Combining Depth Cues 9/30/2008 1 Introduction to Space Perception Realism:
More informationA Pattern-Based Approach to. Automated Application Performance Analysis
A Pattern-Based Approach to Automated Application Performance Analysis Nikhil Bhatia, Shirley Moore, Felix Wolf, and Jack Dongarra Innovative Computing Laboratory University of Tennessee (bhatia, shirley,
More informationNVIDIA GeForce GTX 580 GPU Datasheet
NVIDIA GeForce GTX 580 GPU Datasheet NVIDIA GeForce GTX 580 GPU Datasheet 3D Graphics Full Microsoft DirectX 11 Shader Model 5.0 support: o NVIDIA PolyMorph Engine with distributed HW tessellation engines
More informationMean-Shift Tracking with Random Sampling
1 Mean-Shift Tracking with Random Sampling Alex Po Leung, Shaogang Gong Department of Computer Science Queen Mary, University of London, London, E1 4NS Abstract In this work, boosting the efficiency of
More informationMultiresolution 3D Rendering on Mobile Devices
Multiresolution 3D Rendering on Mobile Devices Javier Lluch, Rafa Gaitán, Miguel Escrivá, and Emilio Camahort Computer Graphics Section Departament of Computer Science Polytechnic University of Valencia
More informationHigh-speed image processing algorithms using MMX hardware
High-speed image processing algorithms using MMX hardware J. W. V. Miller and J. Wood The University of Michigan-Dearborn ABSTRACT Low-cost PC-based machine vision systems have become more common due to
More informationShader Model 3.0. Ashu Rege. NVIDIA Developer Technology Group
Shader Model 3.0 Ashu Rege NVIDIA Developer Technology Group Talk Outline Quick Intro GeForce 6 Series (NV4X family) New Vertex Shader Features Vertex Texture Fetch Longer Programs and Dynamic Flow Control
More informationChoosing a Computer for Running SLX, P3D, and P5
Choosing a Computer for Running SLX, P3D, and P5 This paper is based on my experience purchasing a new laptop in January, 2010. I ll lead you through my selection criteria and point you to some on-line
More informationNVIDIA workstation 3D graphics card upgrade options deliver productivity improvements and superior image quality
Hardware Announcement ZG09-0170, dated March 31, 2009 NVIDIA workstation 3D graphics card upgrade options deliver productivity improvements and superior image quality Table of contents 1 At a glance 3
More informationHow To Use An Amd Ramfire R7 With A 4Gb Memory Card With A 2Gb Memory Chip With A 3D Graphics Card With An 8Gb Card With 2Gb Graphics Card (With 2D) And A 2D Video Card With
SAPPHIRE R9 270X 4GB GDDR5 WITH BOOST & OC Specification Display Support Output GPU Video Memory Dimension Software Accessory 3 x Maximum Display Monitor(s) support 1 x HDMI (with 3D) 1 x DisplayPort 1.2
More informationParallel algorithms to a parallel hardware: Designing vision algorithms for a GPU
Parallel algorithms to a parallel hardware: Designing vision algorithms for a GPU Jun-Sik Kim, Myung Hwangbo and Takeo Kanade Robotics Institute Carnegie Mellon University {kimjs,myung,tk}@cs.cmu.edu Abstract
More informationAccelerating Wavelet-Based Video Coding on Graphics Hardware
Wladimir J. van der Laan, Andrei C. Jalba, and Jos B.T.M. Roerdink. Accelerating Wavelet-Based Video Coding on Graphics Hardware using CUDA. In Proc. 6th International Symposium on Image and Signal Processing
More informationA Study on SURF Algorithm and Real-Time Tracking Objects Using Optical Flow
, pp.233-237 http://dx.doi.org/10.14257/astl.2014.51.53 A Study on SURF Algorithm and Real-Time Tracking Objects Using Optical Flow Giwoo Kim 1, Hye-Youn Lim 1 and Dae-Seong Kang 1, 1 Department of electronices
More informationHigh speed 3D capture for Configuration Management DOE SBIR Phase II Paul Banks Paul.banks@tetravue.com
High speed 3D capture for Configuration Management DOE SBIR Phase II Paul Banks Paul.banks@tetravue.com Advanced Methods for Manufacturing Workshop September 29, 2015 1 TetraVue does high resolution 3D
More informationData Visualization Using Hardware Accelerated Spline Interpolation
Data Visualization Using Hardware Accelerated Spline Interpolation Petr Kadlec kadlecp2@fel.cvut.cz Marek Gayer xgayer@fel.cvut.cz Czech Technical University Department of Computer Science and Engineering
More informationCS231M Project Report - Automated Real-Time Face Tracking and Blending
CS231M Project Report - Automated Real-Time Face Tracking and Blending Steven Lee, slee2010@stanford.edu June 6, 2015 1 Introduction Summary statement: The goal of this project is to create an Android
More informationAutomatic Labeling of Lane Markings for Autonomous Vehicles
Automatic Labeling of Lane Markings for Autonomous Vehicles Jeffrey Kiske Stanford University 450 Serra Mall, Stanford, CA 94305 jkiske@stanford.edu 1. Introduction As autonomous vehicles become more popular,
More informationA Learning Based Method for Super-Resolution of Low Resolution Images
A Learning Based Method for Super-Resolution of Low Resolution Images Emre Ugur June 1, 2004 emre.ugur@ceng.metu.edu.tr Abstract The main objective of this project is the study of a learning based method
More informationA Survey of Video Processing with Field Programmable Gate Arrays (FGPA)
A Survey of Video Processing with Field Programmable Gate Arrays (FGPA) Heather Garnell Abstract This paper is a high-level, survey of recent developments in the area of video processing using reconfigurable
More informationIn mathematics, there are four attainment targets: using and applying mathematics; number and algebra; shape, space and measures, and handling data.
MATHEMATICS: THE LEVEL DESCRIPTIONS In mathematics, there are four attainment targets: using and applying mathematics; number and algebra; shape, space and measures, and handling data. Attainment target
More informationMemory Allocation Technique for Segregated Free List Based on Genetic Algorithm
Journal of Al-Nahrain University Vol.15 (2), June, 2012, pp.161-168 Science Memory Allocation Technique for Segregated Free List Based on Genetic Algorithm Manal F. Younis Computer Department, College
More informationTriangulation by Ear Clipping
Triangulation by Ear Clipping David Eberly Geometric Tools, LLC http://www.geometrictools.com/ Copyright c 1998-2016. All Rights Reserved. Created: November 18, 2002 Last Modified: August 16, 2015 Contents
More informationAutomatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 269 Class Project Report
Automatic 3D Reconstruction via Object Detection and 3D Transformable Model Matching CS 69 Class Project Report Junhua Mao and Lunbo Xu University of California, Los Angeles mjhustc@ucla.edu and lunbo
More informationHardware design for ray tracing
Hardware design for ray tracing Jae-sung Yoon Introduction Realtime ray tracing performance has recently been achieved even on single CPU. [Wald et al. 2001, 2002, 2004] However, higher resolutions, complex
More informationA VOXELIZATION BASED MESH GENERATION ALGORITHM FOR NUMERICAL MODELS USED IN FOUNDRY ENGINEERING
METALLURGY AND FOUNDRY ENGINEERING Vol. 38, 2012, No. 1 http://dx.doi.org/10.7494/mafe.2012.38.1.43 Micha³ Szucki *, Józef S. Suchy ** A VOXELIZATION BASED MESH GENERATION ALGORITHM FOR NUMERICAL MODELS
More informationColour Image Segmentation Technique for Screen Printing
60 R.U. Hewage and D.U.J. Sonnadara Department of Physics, University of Colombo, Sri Lanka ABSTRACT Screen-printing is an industry with a large number of applications ranging from printing mobile phone
More informationLecture 15: Hardware Rendering
Lecture 15: Hardware Rendering Fall 2004 Kavita Bala Computer Science Cornell University Announcements Project discussion this week Proposals: Oct 26 Exam moved to Nov 18 (Thursday) Bounding Volume vs.
More informationSpeed Performance Improvement of Vehicle Blob Tracking System
Speed Performance Improvement of Vehicle Blob Tracking System Sung Chun Lee and Ram Nevatia University of Southern California, Los Angeles, CA 90089, USA sungchun@usc.edu, nevatia@usc.edu Abstract. A speed
More informationGlencoe. correlated to SOUTH CAROLINA MATH CURRICULUM STANDARDS GRADE 6 3-3, 5-8 8-4, 8-7 1-6, 4-9
Glencoe correlated to SOUTH CAROLINA MATH CURRICULUM STANDARDS GRADE 6 STANDARDS 6-8 Number and Operations (NO) Standard I. Understand numbers, ways of representing numbers, relationships among numbers,
More information